California AI Race Locks Out Grassroots Rivals as US Oversight Lags

Panel 4 - 1: Algorithmic Campaigning: A Case Study of California's 2026 Gubernatorial Race — Ai4

Ai4

Surging capital requirements for proprietary AI targeting tools consolidate power among wealthy incumbents while fragmented regulators fail to police algorithmic voter manipulation.

AI-assisted summary of Panel 4 - 1: Algorithmic Campaigning: A Case Study of California's 2026 Gubernatorial Race — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published on YouTube by Ai4 on August 18, 2026, this presentation features Professor Anna Ni of California State University San Bernardino examining the transformation of political campaigns through artificial intelligence. Grounded in the context of California's 2026 gubernatorial race, the session presents an Algorithmic Campaigning Framework to explain how big data, predictive analytics, and automated optimization impact democratic deliberation, transparency, and public policy.

## TL;DR The talk examines the technological shift from early digital outreach to modern AI-driven algorithmic campaigning, using California's 2026 gubernatorial election as an empirical case study. Ni introduces a Six-Component Framework illustrating how voter datafication, predictive analytics, microtargeting, platform amplification, generative content, and continuous feedback loops create an adaptive persuasion engine. The presentation highlights acute democratic dangers, including public sphere fragmentation, psychological manipulation, and surging campaign costs that favor wealthy incumbents over grassroots candidates. Analyzing the fragmented regulatory landscape across federal agencies and state laws facing First Amendment hurdles, the presentation outlines a national governance roadmap centered on ad archives, data broker regulation, and expanded empirical research.

## Atmosphere & style The session carries the crisp, focused cadence of a rigorous academic conference briefing, structuring complex socio-technical systems into clear diagrams and theoretical lenses. The delivery is analytical and measured, yet conveys a steady undercurrent of urgency regarding the unchecked societal risks of what is termed a technological 'big monster.' Listening feels like attending an information-dense masterclass at the intersection of technology, law, and politics, moving briskly through policy friction points to conclude neatly against a tight presentation countdown.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) Modern political campaigns have evolved from simple digital outreach into autonomous, AI-driven algorithmic ecosystems that continuously harvest data and optimize voter persuasion in real time. - [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) Understanding algorithmic campaigning requires four theoretical lenses: deliberative democracy, algorithmic governance, surveillance capitalism, and algorithm-mediated political communication. - [04:23](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=263s) California's 2026 gubernatorial race serves as a prime case study due to high campaign spending, proximity to tech hubs, and early state-level AI legislative interventions. - [08:31](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=511s) The Six-Component Algorithmic Campaigning Framework connects electorate datafication, predictive modeling, microtargeting, platform amplification, and generative AI into a continuous real-time feedback loop. - [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) Algorithmic curation fragments the democratic public sphere into hyper-segmented echo chambers, depriving citizens of a shared factual baseline necessary for collective deliberation. - [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) Federal oversight is severely fractured across the FEC, FCC, and FTC, leaving political data brokers and algorithm-driven campaign architectures largely unregulated. - [15:00](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=900s) California's legislative efforts to regulate deceptive election AI (such as AB 2839 and AB 2655) face substantial enforcement barriers due to First Amendment protections for political speech. - [15:48](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=948s) Campaign tech vendors and consultancies frequently block transparency and algorithmic auditing by claiming their targeting models are proprietary trade secrets. - [15:48](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=948s) While policy debates focus largely on overt deepfakes, subtle and unregulated AI-driven micro-persuasion poses a deeper systemic threat to voter autonomy. - [16:59](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1019s) Effective AI governance requires mandatory public ad archives, oversight of commercial voter data brokers, and unified federal coordination to replace fragmented state laws. - [18:26](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1106s) The steep financial and technical costs of advanced AI campaign tools widen political inequality, entrenching well-funded incumbents while raising insurmountable barriers for grassroots candidates.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) Introduction to Algorithmic Campaigning and Research Foundations

Over the past quarter-century, political campaigns have transformed fundamentally from early digital outreach mechanisms—like sending campaign emails and publishing candidate websites—into sophisticated, AI-enabled algorithmic ecosystems. Modern campaigns operate in an environment where continuous voter data collection, predictive behavioral modeling, and automated optimization dictate strategy in real time. This rapid evolution introduces profound challenges for governance and democratic deliberation, raising fundamental questions about transparency, privacy, and institutional accountability.

To explore this transformation, the presentation examines the technological progression across four core theoretical lenses:

- **Deliberative Democracy:** Normative democratic theory posits that legitimate governance requires a shared public sphere where citizens engage with diverse perspectives and participate in reasoned debate. When algorithmic systems fragment this sphere into narrow echo chambers, bias deepens and the democratic legitimacy of decision-making erodes. - **Algorithmic Governance:** Examines how automated systems control the flow of political information, structure decision-making, and operate without the necessary auditing mechanisms. - **Surveillance Capitalism:** Highlights how voters' behavioral and consumer data are harvested and converted into political assets, shifting value extraction from purely commercial advertising to direct electoral influence. - **Political Communication:** Traces how mass communication has transitioned into an "algorithm-mediated communication system" designed to analyze, persuade, and mobilize voters at scale.

---

[04:23](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=263s) The Evolution of Campaign Tech and the California Case Study

Political campaign technology has progressed through several defining milestones. The 2000 presidential cycle marked the initial shift into the digital realm through early websites, discussion forums, and direct emails. By 2008, Barack Obama's presidential campaign demonstrated the power of social media to mobilize younger demographics and reshape electoral outcomes. By the 2024 cycle and looking toward 2026, campaigns entered the era of true algorithmic campaigning, driven by large language models (LLMs), generative AI tools, and advanced voter analytics platforms.

California's gubernatorial race serves as a primary case study due to its status as a technological hub, its massive electorate, and its reputation for running some of the most expensive political campaigns in history. Furthermore, California has been at the forefront of enacting state-level AI regulations.

```mermaid flowchart LR accTitle: Evolution of Campaign Technologies accDescr: Timeline showing the progression of political campaign tools from 2000 to 2026. subgraph Progression ["Technological Evolution (2000–2026)"] Y2000["2000: Digital Outreach<br/>Websites, emails, and discussion boards"] Y2008["2008: Social Media Era<br/>Social platforms, youth engagement, mobilization"] Y2024["2024–2026: Algorithmic Era<br/>LLMs, predictive analytics, autonomous ad optimization"] Y2000 --> Y2008 Y2008 --> Y2024 end ```

In the California race, several algorithmic campaign tactics were deployed at an unprecedented scale: - **Algorithmic Ad Delivery:** Automated ad networks determine which voters receive specific messages, optimizing timing and exposure frequency based on predictive preference modeling. - **Enhanced Microtargeting:** Data brokers and campaign teams integrate multiple commercial, voter, and behavioral datasets to construct detailed voter profiles. - **Influencer Mobilization:** Campaigns partner with trusted content creators across YouTube, TikTok, and Instagram to distribute targeted political messages through organic channels. - **AI-Assisted Content Generation:** While strict state legislation curbed direct deceptive deepfakes in official campaign ads, generative AI was extensively used behind the scenes for automated email drafting, copywriting, and rapid content iteration.

---

[08:31](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=511s) The Six-Component Algorithmic Campaigning Framework

Drawing on both empirical observations and theoretical literature, an integrated conceptual framework explains how modern algorithmic campaigns operate as self-reinforcing, data-driven feedback systems. Rather than relying on static broadcast messaging, the ecosystem functions as an adaptive machine that continuously senses and responds to voter inputs.

```mermaid flowchart TD accTitle: Six-Component Algorithmic Campaigning Framework accDescr: Diagram illustrating the six interconnected components of the algorithmic campaigning framework. subgraph DataEngine ["1. Data Infrastructure & Analytics"] A["1. Datafication of the Electorate<br/>(Voter, consumer, and behavioral profiles)"] B["2. Predictive Modeling & Analytics<br/>(Behavior forecasting and propensity scoring)"] A --> B end

subgraph TargetingEngine ["2. Delivery & Automation"] C["3. Algorithmic Personalization & Microtargeting<br/>(Tailored messaging for narrow segments)"] D["4. Platform-Mediated Amplification<br/>(Algorithmic reach, boosting, and suppression)"] E["5. AI-Assisted Content Creation & Automation<br/>(Generative copy, visual assets, and deployment)"] B --> C C --> D C --> E end

subgraph FeedbackEngine ["3. Continuous Optimization"] F["6. Continuous Data & Feedback Loop<br/>(Real-time voter reaction and model retraining)"] D --> F E --> F F --> A end ```

The six structural components of this framework include: - **Datafication of the Electorate:** Individual voters are converted into digital dossiers that combine traditional voter files with consumer purchase histories, browsing habits, and behavioral patterns. - **Predictive Modeling and Analytics:** Advanced statistical and machine learning models process aggregated data to forecast individual turnout likelihood and issue receptivity. - **Algorithmic Personalization and Microtargeting:** Specialized algorithms segment audiences and deliver tailored appeals to highly specific voter cohorts. - **Platform-Mediated Amplification:** Social media algorithms decide which messages to boost or suppress based on engagement metrics and audience responses. - **AI-Assisted Content Creation and Automation:** Generative tools rapidly produce and deploy customized campaign communications across multiple channels at exponential scale. - **Continuous Data and Feedback Loop:** The overarching mechanism that transforms the entire framework into a dynamic, "learning system" that refines its persuasion models in real time based on audience interactions.

---

[11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) Systemic Risks, Public Sphere Fragmentation, and Regulatory Gaps

While algorithmic campaigning offers campaign teams significant efficiencies, improved reach, and personalized messaging, it creates severe societal and democratic externalities. Unlike algorithmic systems deployed in regulated industries like finance or healthcare, political algorithms operate largely unaudited and in complete opacity, raising acute concerns about unaccountable bias, data manipulation, and covert favoritism.

The fragmentation of the public sphere stands as the most critical threat. When algorithmic delivery engines curate customized realities for individual voters, citizens no longer inhabit a shared deliberative forum with common facts. This hyper-segmentation accelerates political polarization and fuels the rapid dissemination of AI-generated misinformation.

```mermaid flowchart LR accTitle: Fragmented US Regulatory Oversight accDescr: Chart mapping out the distinct regulatory domains of the FEC, FCC, FTC, and State Laws in political campaigning. subgraph FederalOversight ["Federal Regulatory Landscape"] FEC["FEC<br/>Campaign finance & expenditure disclosures"] FCC["FCC<br/>Broadcast media, robocalls & AI voice disclosures"] FTC["FTC<br/>Commercial data privacy & data broker oversight"] end subgraph StateOversight ["State-Level Action"] States["31+ States (incl. California)<br/>Enacting targeted bans on deceptive political deepfakes"] end ```

The regulatory response across the United States remains highly fragmented and ill-equipped to oversee algorithmic systems: - **Federal Election Commission (FEC):** Focuses narrowly on traditional campaign finance reporting and source disclosures, leaving data-driven algorithmic practices largely untouched. - **Federal Communications Commission (FCC):** Governs broadcast regulations and AI-generated robocalls, but possesses limited jurisdiction over algorithmic social platforms. - **Federal Trade Commission (FTC):** Oversees commercial consumer data protections, yet struggles to regulate data brokers whose information is integrated into political campaign systems. - **State-Level Initiatives:** Over 31 states have enacted laws addressing AI-generated deceptive deepfakes in political campaigns, with California leading the passage of major legislative packages to regulate synthetic campaign media.

[15:00](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=900s) Legislative Pushback: Assembly Bills and First Amendment Tensions

Efforts by state lawmakers to rein in digital election interference have begun targeting both the generation and distribution of deceptive political content. California’s legislative agenda has focused on mandating disclosures for AI-generated or AI-delivered campaign advertisements, outlawing the knowing distribution of deceptive election materials, and imposing explicit monitoring duties on large online platforms. Yet, putting these statutes into practice has sparked immediate constitutional battles. Measures like California Assembly Bill 2839 (AB 2839) and Assembly Bill 2655 (AB 2655) have collided directly with federal jurisprudence safeguarding First Amendment speech protections, rendering actual enforcement a precarious and contested legal issue.

Because these initiatives are pursued piecemeal at the state level rather than through a cohesive federal doctrine, the emerging regulatory landscape remains fractured and riddled with loopholes. Lawmakers face the complex challenge of curbing malicious computational manipulation without infringing upon fundamental constitutional guarantees of political speech.

- **AB 2839 & AB 2655:** State measures designed to restrict deceptive AI-generated election content and assign accountability to major digital distribution platforms. - **Constitutional Friction:** Implementation has stalled or faced legal pushback due to collisions with First Amendment standards protecting core political discourse.

---

[15:48](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=948s) Four Critical Regulatory Gaps in Algorithmic Campaigning

Examining the current oversight landscape reveals at least four structural gaps that prevent meaningful governance of tech-driven political outreach. The first is a severely fragmented regulatory framework; there is no comprehensive, overarching law governing how political campaigns handle constituent data or deploy automated outreach systems. The second gap centers on deficient transparency standards. Meaningful algorithm disclosure is virtually non-existent, as campaign tech vendors and consultancy firms shield their targeting mechanisms under the legal defense that codebases and predictive models are privately held and proprietary trade secrets.

```mermaid flowchart TD accTitle: Four Regulatory Gaps in Algorithmic Campaigning subgraph Gaps["Governance Gaps"] G1["Fragmented Regulatory Framework<br>(No overarching data or campaign law)"] G2["Limited Transparency<br>(Algorithms shielded as proprietary secrets)"] G3["Ungoverned AI-Assisted Persuasion<br>(Micro-targeted cognitive manipulation)"] G4["Deficient Algorithmic Oversight<br>(Unmonitored automated deployment)"] end subgraph Impact["Democratic Impact"] I1["Erosion of Democratic Deliberation"] I2["Heightened Political Manipulation"] end Gaps --> Impact ```

The third and fourth gaps involve the subtleties of machine-driven influence. While overt deepfakes have captured public attention and made their way onto legislative dockets, subtle, AI-assisted persuasion remains dangerously unregulated. These adaptive micro-targeting methods can quietly distort political reality, leaving voters susceptible to psychological manipulation that erodes legitimate democratic processes. Compounding this is an overall absence of institutional oversight into how algorithmic campaigns operate in real time behind closed doors.

- **No Overarching Political Data Law:** Commercial and campaign data collection operate without an integrated legal shield protecting voter information. - **Proprietary Secrecy:** Demands for algorithmic auditing are regularly turned away because proprietary IP rights shield vendors from public disclosure. - **Beyond Deepfakes:** Policy debates focus on forged media while overlooking automated, individualized persuasion architectures designed to nudge voter psychology.

---

[16:59](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1019s) Roadmap for AI Governance: Taming the "Big Monster"

Tackling what can aptly be called "this big monster created by technology advancement" requires an aggressive, multi-pronged policy roadmap rather than ad-hoc band-aids. The first priority is mandating robust transparency and accountability mechanisms, starting with public, searchable archives of all algorithmic campaign advertisements and automated messaging. Second, policymakers must establish clear guardrails specifically tailored to governing AI in political campaigns, with an acute focus on unregulated political data brokerage—an industry that harvests and sells hyper-granular behavioral profiles with virtually no civic oversight.

Equally vital is protecting the democratic public sphere from becoming completely balkanized. When algorithmic curation fractures the electorate into disconnected echo chambers, citizens lose access to a baseline of shared facts. Genuine self-governance relies on the ability of voters to engage in reasoned public deliberation based on common information to take legitimate collective action. Finally, closing these loopholes requires systematic national and federal regulatory coordination, steering away from divergent, state-by-state measures toward a unified, nationwide governance framework.

- **Ad Archiving & Auditing:** Enforcing comprehensive digital repositories for all campaign ads to trace automated distribution, spending, and source targeting. - **Voter Privacy & Brokerage Oversight:** Curtailing the unconstrained commercial buying and selling of citizen psychological profiles for electioneering. - **Rebuilding a Shared Public Sphere:** Creating structural safeguards to counteract fragmented information silos and foster shared democratic deliberation. - **National Legislative Coordination:** Transitioning from inconsistent state bills toward synchronized federal oversight.

---

[18:26](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1106s) Democratic Consequences, Campaign Equality, and Future Research

Algorithmic campaigning is permanently restructuring democratic communication, electoral competition, and political strategy. Beyond the pressing concerns of transparency, voter privacy, and deliberative integrity, this shift creates severe inequalities in political access. As demonstrated in high-stakes environments like California, deploying cutting-edge AI architectures, proprietary datasets, and predictive modeling makes campaigning exceedingly expensive. These surging capital requirements erect prohibitive entry barriers, consolidating power among well-funded political incumbents and deep-pocketed donors while shutting out grassroots challengers.

Because current inquiries have largely established conceptual and structural frameworks, there remains a pressing need for empirical research. Scholars and watchdogs must dive deeper into measurable field evidence to gauge how algorithmic campaigning concretely alters voter turnout, accelerates ideological polarization, influences institutional trust, and reconfigures citizen participation. As the session winds down with only about thirty seconds left, the talk concludes on the imperative to translate these analytical frameworks into actionable democratic protections.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) Over the past quarter-century, political campaigns have transformed fundamentally from early digital outreach mechanisms—like sending campaign emails and publishing candidate websites—into sophisticated, AI-enabled algorithmic ecosystems. — Anna Ni

> [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) When algorithmic systems fragment this sphere into narrow echo chambers, bias deepens and the democratic legitimacy of decision-making erodes. — Anna Ni

> [08:31](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=511s) Rather than relying on static broadcast messaging, the ecosystem functions as an adaptive machine that continuously senses and responds to voter inputs. — Anna Ni

> [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) Unlike algorithmic systems deployed in regulated industries like finance or healthcare, political algorithms operate largely unaudited and in complete opacity, raising acute concerns about unaccountable bias, data manipulation, and covert favoritism. — Anna Ni

> [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) When algorithmic delivery engines curate customized realities for individual voters, citizens no longer inhabit a shared deliberative forum with common facts. — Anna Ni

> [15:48](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=948s) Meaningful algorithm disclosure is virtually non-existent, as campaign tech vendors and consultancy firms shield their targeting mechanisms under the legal defense that codebases and predictive models are privately held and proprietary trade secrets. — Anna Ni

> [15:48](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=948s) While overt deepfakes have captured public attention and made their way onto legislative dockets, subtle, AI-assisted persuasion remains dangerously unregulated. — Anna Ni

> [16:59](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1019s) this big monster created by technology advancement — Anna Ni

> [16:59](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1019s) Genuine self-governance relies on the ability of voters to engage in reasoned public deliberation based on common information to take legitimate collective action. — Anna Ni

> [18:26](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1106s) These surging capital requirements erect prohibitive entry barriers, consolidating power among well-funded political incumbents and deep-pocketed donors while shutting out grassroots challengers. — Anna Ni

## Mentioned Anna Ni, California State University San Bernardino, Barack Obama, Assembly Bill 2839 (AB 2839), Assembly Bill 2655 (AB 2655), Federal Election Commission (FEC), Federal Communications Commission (FCC), Federal Trade Commission (FTC), YouTube, TikTok, Instagram, First Amendment

## Video description Anna Ni, Professor, California State University San Bernardino

How is AI transforming political campaigns, and what does that mean for democracy? In this presentation, I will examine the evolution of campaign technologies from digital outreach to AI-enabled algorithmic campaigning, using the 2026 California gubernatorial race as a case study. I will introduce an Algorithmic Campaigning Framework to explain how big data, predictive analytics, algorithms, and AI are reshaping political communication and campaign strategies while raising new challenges for transparency, privacy, accountability, and democratic deliberation. I will also discuss emerging regulatory responses and identify key governance gaps that policymakers must address. Attendees will gain a practical framework for understanding the intersection of AI, politics, and public policy, and why responsible AI governance is essential for protecting democratic values.

[Source](https://www.youtube.com/watch?v=Eu_LK49dhq4)

中文简报

美监管滞后之际 加州AI竞选将草根对手拒之门外 专有AI定向工具日益高昂的资金门槛巩固了富裕现任者的权力，而四分五裂的监管体系未能遏制针对选民的算法操纵。 本次演讲以加州2026年州长选举为实证案例研究，探讨了从早期数字触达向现代AI驱动型算法竞选的技术转变。Ni提出了一个“六要素框架”，阐明了选民数据化、预测分析、微定向、平台放大、生成式内容以及持续反馈回路如何构建出一个自适应说服引擎。演讲重点指出了严峻的民主危机，包括公共领域碎片化、心理操控，以及有利于富裕现任者却不利于草根候选人的竞选成本飙升。通过分析联邦机构割裂的监管格局及面临《第一修正案》障碍的州级立法，演讲勾勒出了一份以广告档案库、数据经纪商监管和扩大实证研究为核心的国家治理路线图。 现代政治竞选已从早期的数字宣传演变为基于 AI 的算法竞选系统，依托数据化选民档案、预测建模和生成式内容构建起实时反馈闭环。 (4:23) 加州 2026 年州长竞选成为算法竞选的重要案例，高昂的竞选资金与科技枢纽优势推动了算法广告投放与超级微观定向（microtargeting）的大规模应用。 (8:31) 算法策展将民主公共领域割裂为高度分化的信息茧房，剥夺了公民进行集体审议所需的共享事实基础，同时加剧了政治极化。 (11:27) 美国联邦层面在 FEC、FCC 和 FTC 之间的监管碎片化严重，而加州针对深度伪造的立法（如 AB 2839 和 AB 2655）也面临第一修正案宪法保护带来的执法障碍。 (15:00) 算法竞选的高额研发与数据成本大幅推高了参选门槛，权力进一步向资金雄厚的现任者集中，未来需建立广告归档及统一的全国性监管框架。 (18:26)

Reading summary

The presentation argues that modern political campaigning has evolved into an adaptive, algorithmic feedback loop powered by predictive analytics, behavioral data, and generative artificial intelligence. Examining California's 2026 gubernatorial contest as a benchmark, the speaker asserts that these technologies threaten democratic deliberation, reinforce electoral inequality, and outpace fragmented regulatory structures. The Architecture of Closed-Loop Algorithmic Campaigns (8:31) Campaign technology has progressed beyond the basic digital outreach of the 2000 election and the social mobilization of 2008 into an integrated, six-component algorithmic engine. The ecosystem begins with the datafication of the electorate, combining official voter records with commercial consumer histories and web behavior. Predictive models then score individual voters on turnout probability and policy receptivity. These predictive insights feed automated microtargeting, platform-driven amplification, and generative AI content creation, including rapid email drafting and copy generation. Rather than broadcasting static messaging, the infrastructure operates as an autonomous learning system: real-time voter reactions and engagement metrics are continuously fed back into the models to refine persuasion strategies and optimize ad delivery. Public Sphere Fragmentation Beyond Deepfake Distortions (11:27) Applying frameworks from deliberative democracy and surveillance capitalism, the presentation highlights how algorithmic campaigning erodes the shared factual foundation required for legitimate self-governance. When automated platforms deliver bespoke realities to micro-segmented audiences, voters are pushed into isolated echo chambers that accelerate ideological polarization. While public debate and state legislatures have focused primarily on overt synthetic media such as deceptive deepfakes, the source identifies subtle, AI-assisted persuasion as an equally significant democratic threat. Microtargeted cognitive appeals can quietly manipulate voter attitudes without entering the public view, remaining invisible to opposing candidates, journalists, and civic watchdogs. Regulatory Gaps and Constitutional Enforcement Hurdles (15:00) Governance of political technology in the United States remains severely disjointed. The Federal Election Commission oversees narrow financial expenditures, the Federal Communications Commission regulates broadcast media and robocalls rather than social platforms, and the Federal Trade Commission struggles to supervise commercial data brokers feeding campaign machines. Simultaneously, private tech vendors shield targeting algorithms from civic audits under proprietary trade-secret claims. State-level interventions have attempted to close these loopholes, with over 31 states passing measures on deceptive campaign media. California led with bills such as AB 2839 and AB 2655 to penalize deceptive synthetic content and compel platform monitoring. However, these enforcement efforts have stalled amid First Amendment challenges safeguarding core political speech, demonstrating the difficulty of policing computational influence without coherent federal standards. Campaign Inequality and the AI Governance Roadmap (16:59) The capital-intensive nature of cutting-edge algorithmic campaigning creates severe structural imbalances. Deploying proprietary datasets, advanced machine learning, and multi-channel generative tools requires vast financial resources, entrenching well-funded political incumbents and affluent donors while effectively pricing out grassroots challengers. To manage this expanding technological footprint, the speaker proposes a four-part policy roadmap: mandating comprehensive, searchable public archives for algorithmic political ads; establishing privacy oversight over political data brokers; building institutional safeguards to protect common democratic deliberation; and replacing fragmented state bills with synchronized federal oversight. While analytical models successfully map how algorithmic campaigns operate, the presentation concludes that researchers must urgently gather empirical field evidence to measure how these automated systems tangibly alter voter turnout, institutional trust, and civic participation.

精炼阅读摘要

本场研讨指出，现代政治竞选已从早期的数字化联络手段演进为由人工智能与预测模型驱动的闭环算法生态系统；这种以加利福尼亚州2026年州长竞选为代表的技术变革，不仅通过信息分化侵蚀了审议民主的公共基础，还因高昂的资本门槛加剧了选举权力的不平等，亟需跨越州立碎片化监管的系统性治理框架。 技术演进与加州竞选的算法实践 (4:23) 从技术发展脉络来看，政治竞选历经了数个关键节点：2000年大选以候选人网站和直邮开启数字探索；2008年奥巴马竞选团队通过社交媒体动员年轻选民重构选举格局；至2024年乃至2026年周期，大型语言模型（LLM）、生成式AI工具与高级选民画像分析平台推动竞选正式步入真正的算法时代。加州因其庞大的选民体量、科技枢纽地位以及历史上极其昂贵的竞选投入，成为这一技术跃迁的核心实证场域。 在加州州长竞选案例中，算法手段呈现出规模化应用特征：自动化广告网络能够依托预测性偏好模型，在精确计算的时间点向特定选民定向推送议题信息；数据经纪商与竞选团队高度整合商业消费、网络浏览及选民投票历史，构建精细的多维数字档案；竞选团队与YouTube、TikTok和Instagram等平台创作者合作，以原生内容形式渗透细分圈层；在加州针对深度伪造（deepfake）出台严厉法规的约束下，生成式AI被广泛转向幕后的文案编写、筹款邮件起草和极速内容迭代。 六要素自适应算法竞选框架 (8:31) 报告提出了一个由六大相互关联要素构成的概念框架，阐释现代算法竞选如何运作：与传统单向广播式信息传播不同，现代竞选体系是一个能够持续感知并响应选民动态的自适应机器学习反馈系统。 该系统第一阶段为数据与分析引擎，包含“选民数据化”（将公民转变为融合消费和行为记录的数字档案）与“预测建模与分析”（利用机器学习预测选民投票意向及对特定议题的敏感度）；第二阶段为投放与自动化引擎，包含“算法个性化与微观定向”（针对狭窄受众群体定制政治诉求）、“平台媒介化放大”（社交媒体算法基于互动指标进行内容助推或压制）以及“AI辅助内容生成与自动化”（指数级快速生成跨渠道定制素材）；第三阶段则是“持续数据与反馈回路”，系统依据选民的实时互动反馈动态更新预测模型，使说服机制形成不断自我强化的闭环。 公共领域碎片化与政治门槛升高 (11:27) 报告从审议民主理论、算法治理与监视资本主义等多重理论视角出发，指出算法竞选正在引发严重的系统性负外部性。最直接的危害在于公共领域的结构性碎片化：当分发算法为每个选民量身定制政治信息时，公民不再共享客观的事实基准，共同讨论与理性辩论的公共空间瓦解为封闭的政治回音室，加速了意识形态极化和AI虚假信息的扩散。 与此同时，技术的高门槛对选举平等构成了深远威胁。在前沿竞选实践中，部署尖端AI架构、购买第三方细粒度行为数据并运行复杂的预测算法，需要极其高昂的资本支出。讲者指出，这种资金密集的军备竞赛建立起了极高的准入门槛，使得巨额政治献金支持的在任者和建制派候选人得以进一步巩固竞争优势，而缺乏雄厚资源的基层挑战者则面临被系统性边缘化的风险。 监管制度漏洞与第一修正案冲突 (15:00) 当前美国的监管体系严重滞后于算法竞选的发展步伐，呈现出显著的机构碎片化：联邦选举委员会（FEC）主要关注传统的竞选资金流向，对算法数据行为基本不予触及；联邦通信委员会（FCC）的管辖权限于广播媒体和AI自动呼叫电话，无法监管社交平台算法；联邦贸易委员会（FTC）虽监管商业数据隐私，却难以有效遏制政治数据经纪商的大规模买卖。尽管已有超过31个州立法针对政治深度伪造，加州也率先推出了AB 2839和AB 2655等法案，试图限制虚假AI内容并压实大平台监管责任，但这些州级尝试随即与联邦第一修正案保护政治言论的宪法判例产生剧烈摩擦，导致法案实施陷入司法争议与执行停滞。 讲者进一步梳理出当前治理面临的四大核心漏洞：其一，缺乏统领性的政治数据隐私法；其二，竞选科技供应商以知识产权和商业机密为由拒绝公开模型，导致算法透明度和公共审计完全缺失；其三，现行政策过度聚焦于显性的深度伪造影像，却严重忽视了隐蔽的微观心理操纵与AI辅助认知说服；其四，缺乏对算法自动化实时部署的独立机构监督。 报告最终呼吁建立多维度的治理路线图，包括强制建立可公开检索的算法广告库、规制政治数据经纪产业，并推动从各自为政的州级立法转向全国协同监管。然而讲座亦承认，目前的研究主要停留在理论框架与定性归纳阶段，由于算法黑箱与竞选保密机制，现场未能展示更多可量化的实证数据；未来亟待开展基于真实选举环境的实证研究，以系统检验算法干预对投票率、制度信任度和选民参与的具体因果效应。

Detailed notes — zh

## Background 本视频由科技会议频道“Ai4”于 2026 年 8 月 18 日发布在 YouTube 平台，主讲人为加利福尼亚州立大学圣贝纳迪诺分校的 Anna Ni 教授。演讲以 2026 年加州州长竞选为核心案例，详细介绍了人工智能、大数据与预测算法如何重构政治传播与竞选策略，并系统阐释了算法竞选框架及其对民主治理带来的严峻监管挑战。 ## TL;DR 演讲梳理了政治竞选从数字化触达迈向 AI 算法竞选的历史演进，并提出了涵盖选民数据化、预测分析、微靶向投放、平台放大、AI 内容创作与持续反馈回路的六要素生态系统。通过剖析加州州长选举中的算法分发与多源数据整合实践，讲者指出了公共领域碎片化、算法黑箱、隐私侵蚀以及推高竞选资金门槛等系统性民主危机。面对加州立法与联邦第一修正案的司法冲突及监管碎片化现状，讲者呼吁建立强制性政治广告存档制度，严打政治数据掮客，并推动国家层面的顶层立法协同。 ## Atmosphere & style 整场演讲洋溢着严谨扎实的学术洞察与高度现实关怀的批判意识，讲者条理清晰地将复杂的跨学科理论与最前沿的选战动态融会贯通。叙述语调冷静沉着且富有警示意味，将迅猛演进的算法技术生动喻为需要制度规制的“大怪兽”，并在时间压缩的尾声展现出幽默从容的互动氛围，带给观众关于技术重构民主基石的深刻紧迫感与反思空间。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) 政治竞选在过去 25 年间经历了从早期数字化触达到由 AI 赋能的“算法竞选”（Algorithmic Campaigning）体系的根本性范式转移。 - [01:57](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=117s) 本研究引入协商民主、算法治理、监控资本主义与政治传播学四大理论视角，审视数据转化为政治资本及公共领域碎片化对民主合法性的侵蚀。 - [05:51](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=351s) 加州州长竞选成为前沿技术与立法碰撞的试验场，集中展现了算法广告分发、跨维度微靶向、网红节点动员及 AI 辅助内容创作等关键实践。 - [08:38](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=518s) 算法竞选已演变为闭环生态系统，涵盖选民数据化、预测分析、算法个性化、平台放大、AI 内容生成以及持续反馈回路六大咬合要素。 - [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) 算法竞选引发了深层系统性危机，包括选民商业与政治隐私遭肆意关联、缺乏事实共识加剧极化、算法缺乏独立审计黑箱运作以及 AI 虚假信息鉴别困难。 - [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) 美国联邦层面监管权责割裂（FEC、FCC 与 FTC 各自为政），地方各州虽有超 30 个出台 AI 深伪法规，但整体治理格局依然高度滞后且碎片化。 - [15:00](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=900s) 加州率先出台的 AB 2839 与 AB 2655 等前沿法案在具体执行中与宪法第一修正案保障的言论自由标准产生正面司法冲突，遭遇落地阻碍。 - [15:00](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=900s) 算法竞选当前面临四大监管真空：缺乏跨层级顶层立法、算法因商业机密保护导致透明度极低、针对 AI 诱导说服监管放任，以及全流程外部审查机制缺位。 - [17:06](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1026s) 治理算法竞选需建立强制性广告公共存档制度、严控政治数据掮客、重塑共享事实基础以捍卫民主审议空间，并强化国家层面的立法协同。 - [18:26](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1106s) 算法与算力依赖导致竞选成本呈指数级攀升，抬高了政治入场门槛并损害政治平等；后续学术研究亟待从概念框架转向对选民极化与信任度的实证检验。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) 算法竞选的兴起与研究背景 政治竞选在过去 25 年中经历了深刻的范式转移，从最初利用电子邮件、静态网站进行直接数字联络的简单模式，逐步演变升级为由人工智能赋能的“算法竞选”（Algorithmic Campaigning）体系。如今的选民正置身于一个全天候持续收集数据、精准预测选民行为、生成高度个性化说服内容并优化数字竞选决策的庞大系统之中。这一转变不仅深刻重塑了政治传播形态，更对民主制度的合法性与公共治理构成了重大挑战。本研究以最新的加利福尼亚州州长竞选为案例，探讨大数据、人工智能与算法如何重塑竞选策略，并系统审视现有的立法举措与治理盲区。 ```mermaid flowchart TD accTitle: 政治竞选技术的演进路径 subgraph S1["早期数字时代 (2000年代)"] A["数字化触达"] --> A1["电子邮件 / 竞选网站 / 论坛"] end subgraph S2["社交网络时代 (2008年)"] B["社交媒体工具"] --> B1["青年选民动员 / 大规模说服"] end subgraph S3["算法与AI时代 (2024年及以后)"] C["算法竞选体系"] --> C1["大语言模型 / 行为预测 / 自动化决策"] end S1 --> S2 --> S3 ``` - **四大核心研究问题**：梳理竞选技术的历史演进、界定大数据与 AI 重塑的算法竞选概念、结合加州案例分析对民主的潜在影响、识别现存政策与监管漏洞。 --- ### [01:57](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=117s) 四大理论视角与技术演进节点 为了系统剖析算法竞选，研究从四个核心理论视角切入：首先是**协商民主**（Deliberative Democracy），强调合法的民主依赖于一个共享的公共领域，让公众接触多元观点并进行理性政治辩论；一旦该领域碎片化，选民陷入信息狭窄的偏见中，将直接损害民主决策的合法性。其次是**算法治理**（Algorithmic Governance），关注算法如何全面接管信息流动与决策制定，亟需建立审计与监督机制。第三是**监控资本主义**（Surveillance Capitalism），揭示选民数据与全方位行为数据高度融合，这些数据不再仅作为商业经济资源，而已转变为决定选战结果的核心政治资本。最后是**政治传播学**（Political Communication），展示了大众传播如何全面演变为受算法中介与调控的自动化系统。 在历史演进的时间线上，有三个标志性节点清晰勾勒出技术对政治的重塑： - **2000 年大选**：电子邮件、竞选官方网站和在线讨论区初露锋芒，开启数字政治传播序幕。 - **2008 年奥巴马竞选**：社交媒体工具被大规模系统化应用，成为触达年轻群体与激发政治动员的决定性转折点。 - **2024 年及以后**：大语言模型（LLM）、高级选民分析平台与生成式 AI 全面渗入竞选底层基础设施，标志着政治正式进入算法竞选时代。 --- ### [05:51](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=351s) 加州州长竞选案例：算法竞选的四种主要实践 加利福尼亚州不仅是全美科技产业的中心和规模最大的选区之一，其近期举行的州长选举也是历史上耗资最昂贵的竞选之一。加州在竞选技术前沿应用与 AI 立法监管两方面均走在最前列，是观察算法竞选机制的绝佳窗口。在这场竞选活动中，四种技术实践虽然并非首次出现，但在应用规模和深度上均达到了前所未有的水平： 1. **算法驱动的广告分发**：平台算法完全掌控了广告展示给谁、在何时展示以及展示频率，系统自动预测潜在选民的政治偏好并实施精准触达。 2. **多源数据整合与增强型微靶向投放**：竞选团队打通多维度数据孤岛，对特定潜在选民群体进行颗粒度极细的定制化游说。 3. **网红与内容创作者竞选**：竞选团队积极联合 Instagram、YouTube 和 TikTok 上备受信赖的内容创作者（Creators），将其作为信息渗透的关键传播节点。 4. **AI 辅助的内容生成与传播**：受限于加州针对 Deepfake（深度伪造）竞选广告的严格法律法规，候选人较少直接发布全 AI 生成的虚假广告，但大范围使用了 AI 辅助系统来自动撰写竞选邮件、策划宣传文案与分发内容。 --- ### [08:38](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=518s) 算法竞选的六要素生态框架 基于理论推演与实际观察，算法竞选已演变为一个高度闭环、自我进化的动态生态系统。该框架由六个相互咬合的核心组件构成： ```mermaid flowchart TD accTitle: 算法竞选系统六要素框架 subgraph Core["核心驱动"] F["6. 持续数据流与反馈回路 (Continuous Feedback Loop)"] end subgraph Pipeline["算法执行流程"] A["1. 选民数据化<br/>(Datafication of Electorate)"] --> B["2. 预测模型与分析<br/>(Predictive Analytics)"] B --> C["3. 算法个性化与微靶向<br/>(Personalization & Microtargeting)"] C --> D["4. 平台中介的放大/抑制<br/>(Platform Amplification)"] D --> E["5. AI 辅助内容生成与自动化<br/>(AI Content & Automation)"] end E --> F F -.->|实时自我更新| A F -.->|动态参数调优| B F -.->|实时策略优化| C ``` - **选民数据化（Datafication of Electorate）**：选民被完全虚拟抽象为“数据档案”，除传统选民登记信息外，还深度整合了商业消费习惯、网络浏览记录与线下行为轨迹。 - **预测模型与高级分析（Predictive Analytics）**：竞选团队利用预测算法解析选民数据档案，评估其投票倾向与政策敏感点。 - **算法个性化与微靶向（Algorithmic Personalization & Microtargeting）**：利用算法识别重点摇摆人群，定向推送高度定制化的政治信息。 - **平台中介的放大机制（Platform-Mediated Amplification）**：数字平台依据用户实时互动反馈，自主决定对特定政治言论进行流量放大或限流抑制。 - **AI 辅助内容创作与自动化（AI Content Creation & Automation）**：借助 AI 即时生成海量个性化传播素材，以指数级扩散速度触达目标选民。 - **持续数据流与反馈回路（Continuous Data & Feedback Loop）**：前五个环节产生的选民反应数据源源不断回流，驱动系统不断机器学习、自主调优并实时调整说服策略。 --- ### [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) 民主风险与碎片化的监管格局 算法竞选体系虽在选民触达效率和沟通定制化方面带来了显著优势，但也引发了不可忽视的系统性民主危机与伦理风险： | 核心维度 | 算法竞选带来的负面影响与治理困境 | |:--- |:--- | | **隐私侵犯** | 政治数据经纪商（Data Brokers）缺乏有效监管，商业消费数据与政治档案肆意打通关联。 | | **公共领域碎片化** | 算法决定了个体接收的信息内容、时间与频次，公众失去共享的事实基础，加剧社会偏见与政治极化。 | | **算法黑箱与缺乏审计** | 不同于金融或医疗领域的严格算法合规审查，政治竞选算法完全缺乏独立第三方审计，极易滋生不透明的政治偏向。 | | **AI 虚假信息** | 生成式 AI 内容真伪难辨，界定合法政治表达与恶意误导的界限极其困难。 | 面对严峻风险，美国的监管体系却呈现出高度碎片化且滞后的状态： - **联邦层面权责割裂**： - **联邦选举委员会（FEC）**：职能局限于竞选资金流向审查与信息披露。 - **联邦通信委员会（FCC）**：仅负责传统广播电视、政治广告播发及 AI 自动语音电话（Robocalls）的合规披露。 - **联邦贸易委员会（FTC）**：专注于商业数据保护，对已渗入政治领域的商业数据整合监管乏力。 - **州级立法的率先探索**：全美目前已有约 31 个州出台了针对政治竞选中 AI 深度伪造（Deepfakes）与生成式内容的法律法规，其中加州处于立法最前沿，接连出台了多部针对性法案予以规制。 ### [15:00](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=900s) 加州立法的宪法困境与四大监管真空 加州立法机构近期试图通过立法对技术驱动的竞选进行规范，重点要求披露由 AI 生成或传播的政治广告，限制明知故犯地分发具有欺骗性的选举内容，并明确规定大型网络平台在治理欺骗性选战内容方面的法定义务。然而，这些法案的实际执行却引发了极大争议。以加州众议院法案 AB 2839 和 AB 2655 为例，它们在落地时与保障第一宪法修正案（First Amendment）言论自由的联邦法律标准产生了正面冲突，导致司法执行层面面临重重阻碍。当前的政治竞选监管举措整体呈现出高度碎片化的特征，留下了大量亟待填补的制度盲区。 讲者指出，当前的监管生态至少存在四大核心空白： - **碎片化的监管框架**：缺乏一部通用的顶层法律来统一规范政治竞选活动和政治数据的采集与利用。 - **极度有限的透明度要求**：针对算法披露的要求严重不足，科技公司与竞选团队往往以“私营持有”和“专有商业机密（Proprietary）”为由拒绝公开算法运作机制。 - **缺乏有效约束的 AI 辅助说服（AI-assisted persuasion）**：尽管深度伪造（Deepfake）已被列入立法防御议程，但基于 AI 的深层政治说服与微靶向诱导依然处于放任状态，极易沦为政治操弄的工具，直接侵蚀合法民主的基石。 - **对算法竞选整体监管的缺位**：针对算法化选战全流程的独立外部审计与日常监督机制几乎处于空白状态。 ```mermaid flowchart TD accTitle: 算法竞选的四大监管困境 subgraph Gaps["算法竞选的四大监管真空"] G1["法规碎片化<br/>(缺乏跨层级顶层立法)"] G2["透明度不足<br/>(算法被视为专有商业机密)"] G3["AI 诱导说服失控<br/>(深伪之外的政治操弄无规可依)"] G4["算法竞选监督缺位<br/>(缺乏系统性外部审查)"] end subgraph Conflict["现实执行障碍"] C1["州级法案 (如加州 AB 2839 / AB 2655)"] C2["联邦宪法第一修正案<br/>(言论自由司法审查挑战)"] C1 -.->|"抵触与冲突"| C2 end ``` --- ### [17:06](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1026s) 规制“技术怪兽”：应对算法竞选的治理路径 面对技术进步催生的这一头“大怪兽”（Big Monster），讲者半开玩笑地呼吁监管层必须迅速采取实质行动，并提出了四个关键层面的治理路径。首先是全面强化透明度与问责机制，当务之急是建立强制性的竞选广告公共存档制度（Ad Archiving），提升政治广告资金链与投放轨迹的可见度。其次是严格规范 AI 在竞选中的应用并捍卫选民隐私，特别是针对游离于监管边缘的“政治数据掮客”（Political Data Brokerage），必须建立严格的行业准入与数据审计规程。 再者，必须全力维护民主审议（Democratic Deliberation）的公共空间。当前的算法分发导致了严重的公共领域碎片化，人们生活在各自的信息茧房中。治理体系需要探索如何确保公众能够共享基础事实信息，参与理性的政治讨论，进而形成合法有效的集体行动。最后，讲者强调必须打破各州自行其是、政出多门的现状，建立强有力的国家级监管协调机制，由联邦政府牵头整合零散的立法倡议，形成跨司法管辖区的合力。 | 治理维度 | 核心痛点与风险 | 建议应对举措 | |:--- |:--- |:--- | | **透明度与问责** | 算法黑箱与暗箱广告投放 | 建立强制性的竞选广告公开档案库，提升算法透明度 | | **数据隐私与 AI 规范** | 选民隐私遭无底线挖掘，数据掮客横行 | 将 AI 竞选纳入常态规制，严打非法政治数据经纪与滥用 | | **民主审议空间** | 算法推波助澜导致公共领域碎片化 | 保护共享公共事实基础，保障理性辩论与集体决策 | | **法律协同** | 各州各行其是，法规与联邦法冲突频发 | 推动联邦层面的国家级协调，整合碎片化的法案体系 | --- ### [18:26](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1106s) 民主隐忧与未来研究展望 在演讲的结语部分，讲者高度概括了算法竞选对民主沟通与选举政治带来的全方位重塑。算法化手段所引发的次生危机不仅局限于隐私侵犯、透明度缺失、问责困难和审议受损，更深刻影响了政治平等的根基。随着竞选策略高度依赖庞大的算力、高级算法和海量数据，竞选成本呈指数级攀升——正如加州州长竞选所展示的那样，高昂的资金门槛直接沦为一道壁垒，将许多缺乏雄厚财力支持的普通候选人拒之门外，严重恶化了政治参与的公平性。 回顾本项研究，讲者坦言当前成果主要建立在概念性分析框架之上，尚未完全展开深入的实证量化研究。未来的学术努力应当进一步聚焦于实证检验：深入考察算法竞选究竟在何种程度上改变了选民的真实投票行为，是否直接加剧了政治极化，以及对选民的制度信任和政治参与意愿造成了怎样的长远冲击。 - **竞选准入门槛**：竞选活动的算法化与技术化使得选战耗资极其昂贵，极大地抬高了政治入场门槛，加剧了候选人之间的资源不平等。 - **实证研究转向**：未来研究需从理论架构走向实证数据，系统测度算法竞选对选民行为、社会极化（Polarization）、政治信任度（Trust）以及投票参与率（Participation）的实质影响。 - **尾声互动**：讲者在结束演讲时向现场示意提问时间，笑称时间已严重压缩，仅剩最后 30 秒进行简短收尾。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=4s) 选民正置身于一个全天候持续收集数据、精准预测选民行为、生成高度个性化说服内容并优化数字竞选决策的庞大系统之中。 — Anna Ni > [01:57](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=117s) 合法的民主依赖于一个共享的公共领域，让公众接触多元观点并进行理性政治辩论；一旦该领域碎片化，选民陷入信息狭窄的偏见中，将直接损害民主决策的合法性。 — Anna Ni > [01:57](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=117s) 选民数据与全方位行为数据高度融合，这些数据不再仅作为商业经济资源，而已转变为决定选战结果的核心政治资本。 — Anna Ni > [05:51](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=351s) 平台算法完全掌控了广告展示给谁、在何时展示以及展示频率，系统自动预测潜在选民的政治偏好并实施精准触达。 — Anna Ni > [08:38](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=518s) 选民被完全虚拟抽象为“数据档案”，除传统选民登记信息外，还深度整合了商业消费习惯、网络浏览记录与线下行为轨迹。 — Anna Ni > [11:27](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=687s) 不同于金融或医疗领域的严格算法合规审查，政治竞选算法完全缺乏独立第三方审计，极易滋生不透明的政治偏向。 — Anna Ni > [15:00](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=900s) 尽管深度伪造（Deepfake）已被列入立法防御议程，但基于 AI 的深层政治说服与微靶向诱导依然处于放任状态，极易沦为政治操弄的工具，直接侵蚀合法民主的基石。 — Anna Ni > [17:06](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1026s) 面对技术进步催生的这一头“大怪兽”（Big Monster），监管层必须迅速采取实质行动。 — Anna Ni > [18:26](https://www.youtube.com/watch?v=Eu_LK49dhq4&t=1106s) 竞选策略高度依赖庞大的算力、高级算法和海量数据，竞选成本呈指数级攀升……高昂的资金门槛直接沦为一道壁垒，将许多缺乏雄厚财力支持的普通候选人拒之门外，严重恶化了政治参与的公平性。 — Anna Ni ## Mentioned Anna Ni, 加利福尼亚州立大学圣贝纳迪诺分校 (CSUSB), 2026年加州州长竞选, 奥巴马 (Barack Obama), 联邦选举委员会 (FEC), 联邦通信委员会 (FCC), 联邦贸易委员会 (FTC), 美国宪法第一修正案, 加州众议院法案 AB 2839, 加州众议院法案 AB 2655, Instagram, YouTube, TikTok, 大语言模型 (LLM), 生成式 AI (Generative AI), 深度伪造 (Deepfake) ## Video description Anna Ni, Professor, California State University San Bernardino How is AI transforming political campaigns, and what does that mean for democracy? In this presentation, I will examine the evolution of campaign technologies from digital outreach to AI-enabled algorithmic campaigning, using the 2026 California gubernatorial race as a case study. I will introduce an Algorithmic Campaigning Framework to explain how big data, predictive analytics, algorithms, and AI are reshaping political communication and campaign strategies while raising new challenges for transparency, privacy, accountability, and democratic deliberation. I will also discuss emerging regulatory responses and identify key governance gaps that policymakers must address. Attendees will gain a practical framework for understanding the intersection of AI, politics, and public policy, and why responsible AI governance is essential for protecting democratic values. [Source](https://www.youtube.com/watch?v=Eu_LK49dhq4)

Source quotes

[0:04] Anna Ni: 选民正置身于一个全天候持续收集数据、精准预测选民行为、生成高度个性化说服内容并优化数字竞选决策的庞大系统之中。 [1:57] Anna Ni: 合法的民主依赖于一个共享的公共领域，让公众接触多元观点并进行理性政治辩论；一旦该领域碎片化，选民陷入信息狭窄的偏见中，将直接损害民主决策的合法性。 [1:57] Anna Ni: 选民数据与全方位行为数据高度融合，这些数据不再仅作为商业经济资源，而已转变为决定选战结果的核心政治资本。 [5:51] Anna Ni: 平台算法完全掌控了广告展示给谁、在何时展示以及展示频率，系统自动预测潜在选民的政治偏好并实施精准触达。 [8:38] Anna Ni: 选民被完全虚拟抽象为“数据档案”，除传统选民登记信息外，还深度整合了商业消费习惯、网络浏览记录与线下行为轨迹。 [11:27] Anna Ni: 不同于金融或医疗领域的严格算法合规审查，政治竞选算法完全缺乏独立第三方审计，极易滋生不透明的政治偏向。 [15:00] Anna Ni: 尽管深度伪造（Deepfake）已被列入立法防御议程，但基于 AI 的深层政治说服与微靶向诱导依然处于放任状态，极易沦为政治操弄的工具，直接侵蚀合法民主的基石。 [17:06] Anna Ni: 面对技术进步催生的这一头“大怪兽”（Big Monster），监管层必须迅速采取实质行动。 [18:26] Anna Ni: 竞选策略高度依赖庞大的算力、高级算法和海量数据，竞选成本呈指数级攀升……高昂的资金门槛直接沦为一道壁垒，将许多缺乏雄厚财力支持的普通候选人拒之门外，严重恶化了政治参与的公平性。

Source quotes

[0:04] Anna Ni: Over the past quarter-century, political campaigns have transformed fundamentally from early digital outreach mechanisms—like sending campaign emails and publishing candidate websites—into sophisticated, AI-enabled algorithmic ecosystems. [0:04] Anna Ni: When algorithmic systems fragment this sphere into narrow echo chambers, bias deepens and the democratic legitimacy of decision-making erodes. [8:31] Anna Ni: Rather than relying on static broadcast messaging, the ecosystem functions as an adaptive machine that continuously senses and responds to voter inputs. [11:27] Anna Ni: Unlike algorithmic systems deployed in regulated industries like finance or healthcare, political algorithms operate largely unaudited and in complete opacity, raising acute concerns about unaccountable bias, data manipulation, and covert favoritism. [11:27] Anna Ni: When algorithmic delivery engines curate customized realities for individual voters, citizens no longer inhabit a shared deliberative forum with common facts. [15:48] Anna Ni: Meaningful algorithm disclosure is virtually non-existent, as campaign tech vendors and consultancy firms shield their targeting mechanisms under the legal defense that codebases and predictive models are privately held and proprietary trade secrets. [15:48] Anna Ni: While overt deepfakes have captured public attention and made their way onto legislative dockets, subtle, AI-assisted persuasion remains dangerously unregulated. [16:59] Anna Ni: this big monster created by technology advancement [16:59] Anna Ni: Genuine self-governance relies on the ability of voters to engage in reasoned public deliberation based on common information to take legitimate collective action. [18:26] Anna Ni: These surging capital requirements erect prohibitive entry barriers, consolidating power among well-funded political incumbents and deep-pocketed donors while shutting out grassroots challengers.

Original transcript

[0:04] What I really presents today is emerging  in technology in facing political [0:11] transformation what I call it an algorithmic  campaigning. And uh what what I'm going to use the [0:18] recent most recent California governor's race as a  case to illustrate this concept. And uh I'm going [0:26] to review some of the legislative initiatives and  identify the gaps for governance and democracy. [0:36] Okay, for the past 25 years political campaigns  has evolved from direct digital outreach such as [0:44] using emails and your websites to reaching  voters to AI-enabled algorithmic systems. [0:53] And people are increasingly find that  they are facing the system that constantly [1:00] using, collecting data, predicting voter  behaviors, and create personalized persuasion, [1:07] and also uh making optimized  digital campaign decision system. [1:13] So, this shift has significant implications for  democracy and the governance. And my research [1:21] will focus on these four questions. First of all,  I'm going to have a brief review about how this [1:27] campaign technology has evolved and evolved  into this what I call algorithmic campaigning. [1:34] And secondly, we're going to introduce the concept  and by see how big data AI and algorithm shaping [1:41] the campaigns. And lastly, I'm going to discuss  some of the democratic implications using the [1:47] case of California's recent governor's race  and I identify some of the regulatory gaps. And [1:57] uh my my study based on draws insights from this  four basic theoretical perspective. First of all, [2:05] deliberative democracy lays out the normative  foundation about democracy, which means they [2:12] generally argue scholars argue that in order to  have a legitimate democracy, it's better to have a [2:19] public share public sphere that people are  exposed to diverse views and participate in in [2:27] um participating in meaningful reasoning reasoned  political inter-exchange of arguments. If [2:35] the public of such public sphere is fragmented,  people are exposed to limited arguments. They will [2:44] in they will increase their bias and also hurting  the legitimacy of a democratic decision-making. [2:51] And a second theoretical view is about  algorithmic governance, which basically under [2:59] examining how technology increasingly affecting  the information flow and shaping campaigning, [3:06] decision-making. And from the past the contract  dissemination of the information, those algorithm [3:14] is affecting all the decision-making and require  some oversight of algorithm. And the second the [3:23] thirdly is the surveillance capitalism. This one  basically examine examines that recently you have [3:30] seen political voter data are highly integrated  behavioral data, which means that those data [3:38] is no longer offering just economic resources but  also political values, political resources and [3:47] affecting the overall outcome of the campaign. And  lastly, political communication is a body of study [3:55] examining how political information in the  public affair election election campaigns, [4:02] how the information people are collecting  information, persuading voters, and also [4:09] motivating voters into political actions. And  the this body of study shows you how the massive [4:16] communication now became algorithm mediated the  communication system in political campaigns. [4:23] And here I give you a very brief overview  about the evolution of political campaign [4:29] technologies because time frame time limit, I'm  going to briefly point a few very defining point. [4:36] In the 2000 uh a presentation campaign, we have  seen that digital technology such as websites, [4:45] emails, and the discussion boards is  starting to shaping campaign communication. [4:51] Another very critical point is 2008 and  the Obama campaign that you have seen the [4:57] social media tools have been widely applied  and people found it's a very defining [5:03] technology for political campaign outcomes  because in that campaign those tools exact [5:10] huge effect to reaching younger voters and uh and  the persuading more a mobilization actions and [5:20] affecting campaign results. And lastly, I want  point out the point of 2024, which is that we [5:27] starting to see the presidential campaign the AI  enabled technology appear in in those campaigns. [5:34] AI tools, large language models, and also advanced  voter analytics platforms are shaping a lot of the [5:43] campaign infrastructures and so on. So, we entered  in age what do we call the algorithm campaigning. [5:51] The reason I'm using California as a case study  is mainly because it's mostly recent people [5:57] really fresh and California is a as a very  advanced state with campaigning technology [6:04] development with a large technology sector and  the largest electoral kind of voters in the state [6:12] and this campaign is also marked one of the most  expensive governor's campaign in the history. And [6:20] and and another reason is because California  recently enacted AI legislation regulating [6:26] some of the campaign practices. So, here what  I want to point out to you is that in this [6:34] algorithm campaigning in California's campaign we  have seen several practices. Even though they were [6:40] not new, we have seen them before, but they  are kind of more expanded in larger scale. [6:45] First of all, is algorithmic ad delivery  and the algorithm platforms deciding what [6:53] what kind of customer what people see and when  they see and how frequently see different ads. [7:00] They're predicting who are  those potential voters and [7:05] the political preferences. Those delivery targeted  delivering them. And then we also see people in [7:13] the campaign system integrating data sources and  the more enhanced political microtargeting to [7:21] some potential voters. And then the third  one is this influencer campaigning that [7:29] campaign use trusted creators like  Instagram, like YouTube and and [7:38] TikTok and so on as outlet for for  disseminating information. And lastly, [7:44] even though because of the recent  enactment of the California law against uh [7:50] deep fake of AI-generated campaigning  advertisement, then we didn't see directly those [7:57] those candidates using AI-generated contents, but  there are definitely assisted communication and [8:06] such as email generating, content generating,  and so on. And those practices. So, [8:12] how many of you have seen AI-generated political  campaign advertisement? Yes, I I see. This is a [8:20] very starting to be a very highly regulated  attention of regulations. And so, this is [8:31] the California governor's race. And the drawing  on those theoretical as well as practical [8:38] observations, what I trying to propose is a  framework to understand this campaigning system [8:45] and which has been differentiated different or  became a very integrated ecosystem shaping the [8:53] political sphere of um the United States.  And this conceptual framework essentially [9:02] has six components. First the one is the  datafication of electorate. So, in this regard [9:11] uh information, people are the each voters  became data file and which integrated [9:19] data not just only your voter data, but also your  consumer data, your behavioral data, and they [9:27] they this is a voter became a data file in this  virtual environment. And second component is a [9:33] predictive model and the analytics. Campaigns use  a lot of data analytics to predicting your voter [9:41] behavior by integrating all those information  informations of your data file. The third [9:47] component is the algorithmic personalization  and microtargeting. And the system using [9:56] algorithm to analyze your data predicting behavior  and the targeting who are the preferential voters [10:03] and disseminating information to a specific  groups. So, the the fourth one is the platform [10:11] mediated amplification. The the platform hosting  the information that will define what information [10:18] to amplify, what information to suppress based  on the feedback, based on the voter behavior, [10:25] constant feedback from the system. And  lastly is the AI assisted content creation [10:32] and automation. The the AI enable you instantly  creating multiple personalized communication [10:41] channel to disseminating extremely exponential  scales to reaching out to voters. And this five [10:49] components constantly being informed by  the sixth the sixth component, which is the [10:55] continuous data and the feedback loop that  make this a system that is a learning system [11:03] how do you say constantly updating the system,  learning yourself, reacting to different feedback [11:08] from from the system. Overall, the ecosystem of  campaigning become highly integrated and also [11:18] having a very profound impact of a campaigning  outcome. And based on this understanding, [11:27] what I'm trying to say analyze is what are  the potential risks and benefits. We have seen [11:35] this ecosystem, this political campaign algorithm  campaigning system generate a lot of benefits [11:42] such as improved voter outreach, enhancing the  overall efficiency and communication to voters [11:50] as well as more customized communication. But  however, there is a serious concerns. First of all [11:57] is the privacy concerns and there's very limited  regulation regarding the political data brokerage [12:05] and the linkage of variety data sources also  create a huge privacy concerns. And secondly, [12:14] probably one of the most important  is this fragmented public sphere [12:18] because algorithm now decide who will receive what  information and at what time and at what frequency [12:25] and the people are no longer have a commonly  shared public sphere. So this fragmentation more [12:32] or less will will country more more bias, more  political polarization and and micro targeting [12:41] for for for campaign for information is also  contributing the political polarization. And [12:49] the algorithm of campaign used are rarely under  supervision or audit. It's not like financial [12:56] algorithm or health care algorithm. So this  was largely left unaudited. So that this may [13:07] opaque algorithm may contribute to political bias  or maybe favoritism for some political groups and [13:14] so on. And also AI generated misinformation is a  huge concern and as you see some some of them are [13:22] very fake or Uh, how you define the boundary of  what is separate or what is what is not is a big [13:30] challenging. And those um concerns has been uh  has been through a lot of policy attention. Uh, [13:39] currently however, currently effort in regulating  uh the the campaign system is very limited. [13:47] And overall the US's regulatory landscape is  very fragmented and largely at the federal [13:54] level there are three regulated uh uh bodies. One  is the Federal Election Commission. FEC mainly uh [14:02] regulates campaign finance and disclosure.  And the Federal Communication uh Commission [14:09] uh is regulating broadcasting, political ads,  and AI robot robo uh robocall um uh disclosure uh [14:18] issues. Uh, the FTC, the Trade  Commission, is regulated commercial data [14:23] uh which now is more like integrated  into the campaign political data. So, [14:28] uh and uh regarding to the AI generated content,  there are a lot of initiatives. Overall, I think [14:35] there are 31 currently uh 31 states have enacted  some kind of laws to regulate the deep fakes of AI [14:43] generated content in in uh in political campaigns.  And California is one of the most advanced in in [14:51] practicing and in regulating that. Uh, in 2006 uh  2005 and 6, California passed three Assembly bills [15:00] and uh uh trying to regulate the disclosure of AI  generated well uh or delivered uh political ads [15:08] as well as the restriction for knowingly  distributed deceptive election content as well [15:13] as duties of large online platforms um regarding  deceptive campaign election content. However, [15:21] the implementation of those Assembly bills are  very controversy as of now. AB 2839 and AB 2655 [15:32] are kind of in conflicting with the some of  the federal federal laws safeguarding the [15:39] constitutional standards of First Amendment. So,  the implementation is still a big issue. Some So, [15:48] what we can see is all the initiatives  regulating the political campaign in this [15:53] age is very fragmented and leaving a lot of gaps  to fill. And what I basically think at least there [16:01] are four regulatory gaps. One is the fragmented  regulatory framework. There's not overarching law [16:08] guarding the private guarding political campaign  or or or political data. And the secondly, [16:15] the transparency requirement is very limited.  For for example, algorithm disclosure and it's [16:21] always a lot of say it's a privately held and also  considered proprietary. And thirdly, it's poorly [16:28] governed AI-assisted persuasion. Even though  deepfake has been put into the agenda for for [16:35] safeguard, however, how AI-assisted persuasion can  be a big issue because it can be subject to the [16:41] to a lot of political manipulation. And and the  hurting the legitimate democracy. And like I said, [16:49] there's also very limited oversight of algorithm  campaigning. And so, what I trying to see is that [16:59] there are at least taking some actions towards  regulating this big monster [laughter] created by [17:06] technology advancement. First of all, we need  enhanced transparency accountability, especially [17:13] archiving campaign advertisement and enhancing  the the of this issue. Second is um uh we need to [17:22] implement govern AI in this political  campaign uh and protect the voter privacy [17:29] uh especially the political data brokerage and  this one uh is a very limited political oversight. [17:37] And we also uh safeguard democratic deliberation  especially addressing this issue of fragmented [17:45] public sphere of political deliberation. How we  can ensure uh people sharing common information [17:53] engaging reasoned political deliberation and  uh taking legitimate kind of a collective [18:00] action. And lastly um there are also need a  significant political regulatory coordination. [18:08] Instead of each state, maybe federal  government has driving for different initiative [18:13] addressing different aspects of it. Maybe  uh there are some nation nationwide or [18:19] kind of actions to coordinating all the  legislative initiatives. And so overall, [18:26] what I'm trying to say here uh is that algorithm  campaigning is transforming democratic uh [18:33] communication and electoral politics. And it raise  a lot of concerns about privacy, transparency, [18:40] accountability, uh equality, and deliberation.  I didn't forget to mention about equality. [18:45] The main thing is that because campaign becomes  so pricey, so expensive like the California one, [18:51] and uh it also created a higher entry barriers  for uh politicians to enter into this game. [18:58] And lastly, uh because this research is very  general, it's lay out a more conceptual framework. [19:04] We didn't get into very good empirical uh study  into it. There could be a lot of future effort [19:11] deliberated to examine the fact on  voter behavior, polarization, trust, and [19:19] participation as well. Sorry. So, lastly, if  you have any questions and discussions should [19:25] we hold for later or maybe got 1 minute? Okay.  So, sorry. It's just a 30 seconds. [laughter]